COVID-19

Covid-19 discussion, bring your own statistics
Chris Preston
Stargoon
Posts: 118
Joined: Tue Nov 12, 2019 8:05 am

Re: COVID-19

Post by Chris Preston » Fri Nov 20, 2020 10:24 pm

sTeamTraen wrote:
Fri Nov 20, 2020 7:47 pm
Squeak wrote:
Fri Nov 20, 2020 1:26 am
Holy hell. South Australia locked down because one of the patients lied to contact tracers. This person seemed to be entirely unconnected from the other cases, apart from picking up a takeaway pizza from a place where one of the other patients worked. It turns out he worked several shifts over the past week, so counts as a very close contact. It isn't a super-infectious, scary new strain, just an idiot who lied. :(

Contact tracers were already chasing all the pizza shop's customers but that's a lot of people to track down and test.
Thing is, this is going to happen a lot. Politicians have a model of the population in which mostly middle-class people lead mostly middle-class lives. The people who live on the margins usually get ignored because they don't vote much and if they push things too far, our populist interior minister will lock them up, so they don't threaten us. The virus is a great leveller in this regard.

I remember seeing a documentary some years ago about sex workers in Brazil who didn't use condoms because they could charge something like $5 or $10 extra per time. From their point of view this was entirely rational. Until we are at least able to understand why that might be, we aren't going to be able to get people to comply enough to knock the virus on the head. I'm now at the stage where I assume we will be in permanent yo-yo lockdown until we get 80-90% of the population vaccinated.
I am not going to repeat the current set of rumours, but there seems to be more to this than meets the eye.
Here grows much rhubarb.

Chris Preston
Stargoon
Posts: 118
Joined: Tue Nov 12, 2019 8:05 am

Re: COVID-19

Post by Chris Preston » Sat Nov 21, 2020 9:23 am

Last day of the lock down today, although have been allowed to go out for exercise today. Another 19,000 tests yesterday, one new case. A person already in quarantine. Currently there are more than 5000 close contacts in quarantine.

However, it is back to working from home for me on Monday.
Here grows much rhubarb.

User avatar
Opti
Snowbonk
Posts: 545
Joined: Mon Nov 11, 2019 11:21 pm
Location: Got a home right by the sea.

Re: COVID-19

Post by Opti » Sat Nov 21, 2020 6:55 pm

A good, longstanding, friend of mine (guitarist in my last band in UK) has been getting more and more swayed by f.cking social media nonsense lately. Him and his wife felt a bit ill a week ago ... but left it 5 days to seek a test. The wife works in the local post office/community store in the village. 2 days ago they both tested positive. Right now my friend has been admitted to intensive care.

The whole village is righteously angry at both of them, yet very worried for their welfare. So many people in quarantine. Village life, eh?
I don't miss it at all.
He cannot be killed by conventional weapons. True. He and Mrs Opti are wicked cool.

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Tue Nov 24, 2020 12:50 pm

Here is my plot of the last 20 years updated for week 46:

Image
Have you considered stupidity as an explanation

User avatar
Bird on a Fire
Light of Blast
Posts: 4366
Joined: Fri Oct 11, 2019 5:05 pm
Location: nadir of brie

Re: COVID-19

Post by Bird on a Fire » Tue Nov 24, 2020 1:10 pm

Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
THINK BIG AND UPEND THE SYSTEM

#ShowYourStripes

User avatar
Woodchopper
After Pie
Posts: 2105
Joined: Sat Oct 12, 2019 9:05 am

Re: COVID-19

Post by Woodchopper » Tue Nov 24, 2020 2:37 pm

Bird on a Fire wrote:
Tue Nov 24, 2020 1:10 pm
Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
See here and the follow-up: viewtopic.php?f=8&t=1687&p=48566&#p48566

User avatar
Gfamily
After Pie
Posts: 1890
Joined: Mon Nov 11, 2019 1:00 pm
Location: NW England

Re: COVID-19

Post by Gfamily » Tue Nov 24, 2020 2:38 pm

Bird on a Fire wrote:
Tue Nov 24, 2020 1:10 pm
Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
I don't have figures, but I understand the overall excess death rate quite closely tracks the Covid case rate.

The thought is that deaths due to delayed procedures would have a longer tail, so would (likely) have contributed to delaying the fall off of the excess deaths.
My avatar was a scientific result that was later found to be 'mistaken' - I rarely claim to be 100% correct
ETA 5/8/20: I've been advised that the result was correct, it was the initial interpretation that needed to be withdrawn
Meta? I'd say so!

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Tue Nov 24, 2020 2:57 pm

Gfamily wrote:
Tue Nov 24, 2020 2:38 pm
Bird on a Fire wrote:
Tue Nov 24, 2020 1:10 pm
Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
I don't have figures, but I understand the overall excess death rate quite closely tracks the Covid case rate.

The thought is that deaths due to delayed procedures would have a longer tail, so would (likely) have contributed to delaying the fall off of the excess deaths.
They have done after the first 5-weeks of the peak (LH image - RH image is Excess deaths less COVID-19 deaths)

Image

Again, from the ONS weekly stats for England and Wales

But that's not the same question, because there were reductions in excess deaths due to pollution reduction, for example, and later increases due to the downsides to lockdown as well.
Have you considered stupidity as an explanation

User avatar
Bird on a Fire
Light of Blast
Posts: 4366
Joined: Fri Oct 11, 2019 5:05 pm
Location: nadir of brie

Re: COVID-19

Post by Bird on a Fire » Tue Nov 24, 2020 3:06 pm

Woodchopper wrote:
Tue Nov 24, 2020 2:37 pm
Bird on a Fire wrote:
Tue Nov 24, 2020 1:10 pm
Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
See here and the follow-up: viewtopic.php?f=8&t=1687&p=48566&#p48566
That's very helpful, thanks :)
THINK BIG AND UPEND THE SYSTEM

#ShowYourStripes

User avatar
Woodchopper
After Pie
Posts: 2105
Joined: Sat Oct 12, 2019 9:05 am

Re: COVID-19

Post by Woodchopper » Tue Nov 24, 2020 3:10 pm

Bird on a Fire wrote:
Tue Nov 24, 2020 3:06 pm
Woodchopper wrote:
Tue Nov 24, 2020 2:37 pm
Bird on a Fire wrote:
Tue Nov 24, 2020 1:10 pm
Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
See here and the follow-up: viewtopic.php?f=8&t=1687&p=48566&#p48566
That's very helpful, thanks :)
Get in touch if you want the FT article.

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Tue Nov 24, 2020 3:22 pm

Bird on a Fire wrote:
Tue Nov 24, 2020 1:10 pm
Are there any estimates of the % of excess deaths that aren't covid-related, e.g. from delayed procedures etc?

There's a fairly large gulf between official and excess deaths, and it would be interesting to know which is closer to accurate.
Lots here on positive and negative estimates:

https://www.ons.gov.uk/news/statementsa ... dmorbidity
Approximately 16,000 excess deaths are estimated because of changes in emergency care and social care within a year from March 2020 – the majority of these are deaths in care homes; changes to elective care, primary, and community care are not expected to result in deaths in the short term in this scenario.

Between March 2020 and March 2021, the wider impacts of social distancing measures are estimated to reduce mortality, resulting in 7,000 fewer deaths than expected; the main cause of this change is less air pollution early in the outbreak, though the results of that improvement to air quality are expected to actually occur over the longer term.

Over the whole period presented from March 2020 to more than five years from now, there are approximately 25,000 excess deaths resulting from social distancing and economic impacts in the chosen scenario.

The negative health impacts of social distancing amount to a loss of 88,000 QALYs up to March 2021, so mortality improves over this period but health worsens – the main causes of this are increases in: depression; musculoskeletal disorders, resulting from more people working from home without suitable equipment; and domestic abuse.
Have you considered stupidity as an explanation

User avatar
Bird on a Fire
Light of Blast
Posts: 4366
Joined: Fri Oct 11, 2019 5:05 pm
Location: nadir of brie

Re: COVID-19

Post by Bird on a Fire » Tue Nov 24, 2020 4:05 pm

Ok, so it sounds like most of the non-covid contribution to extra deaths would be expected to contribute to a longer tail, perhaps explaining an increasing number over time, whereas probably the initial spike in deaths was mostly undiagnosed covid?

Full disclosure: I'm on an ecological population modelling course this week, which is largely focussed on separating out variance arising from the observation process from the underlying biological process. The kinds of patchy/biased/incomplete detections of covid we have from the early outbreaks are very similar to the kinds of patchy/biased/incomplete data ecologists have to work with when trying to count whales or whatever, and - it seems to me - probably susbtantially more patchy/biased/incomplete than epidemiologists generally have to work with. So there's this idea rattling around my head of trying some ultracrepidarian carpet-bagging (largely for my own amusement) of trying to model the covid first wave by explicitly incorporating detection processes via some sort of hidden Markov model. I think it would work but I don't really have time ;)
THINK BIG AND UPEND THE SYSTEM

#ShowYourStripes

User avatar
sTeamTraen
Dorkwood
Posts: 1211
Joined: Mon Nov 11, 2019 4:24 pm
Location: Palma de Mallorca, Spain

Re: COVID-19

Post by sTeamTraen » Tue Nov 24, 2020 5:18 pm

A tale of two numbers in the UK figures today. 11299 is the lowest daily number of cases since 3 October, while 608 is the highest daily number of deaths since 13 May. In summary, the lockdown is having an effect, but another 10,000-15,000 deaths are baked in.

One possible small piece of good news is that Tuesdays typically have the highest numbers (probably reflecting a full day on Monday, plus odds and ends trickling in from the weekend). The day of the week with the closest to the average is Saturday.
Something something hammer something something nail

User avatar
shpalman
After Pie
Posts: 2290
Joined: Mon Nov 11, 2019 12:53 pm
Location: One step beyond

Re: COVID-19

Post by shpalman » Tue Nov 24, 2020 6:00 pm

https://twitter.com/healthdpt/status/13 ... 18029?s=19

There might be an issue with the Pillar 2 numbers, in Northern Ireland at least.
molto tricky

AMS
Snowbonk
Posts: 420
Joined: Mon Nov 11, 2019 11:14 pm

Re: COVID-19

Post by AMS » Tue Nov 24, 2020 6:46 pm

Don't know if this is covered in any of the linked articles, but some non-covid excess deaths will take a while to actually happen if they result from missed diagnoses. In particular, cancer outcomes are generally better when it is detected early - can't remember where I saw the numbers, but stuff like colon cancer diagnosis rates are way down this year, which will translate into more deaths over the next few years.

User avatar
sTeamTraen
Dorkwood
Posts: 1211
Joined: Mon Nov 11, 2019 4:24 pm
Location: Palma de Mallorca, Spain

Re: COVID-19

Post by sTeamTraen » Tue Nov 24, 2020 7:11 pm

AMS wrote:
Tue Nov 24, 2020 6:46 pm
Don't know if this is covered in any of the linked articles, but some non-covid excess deaths will take a while to actually happen if they result from missed diagnoses. In particular, cancer outcomes are generally better when it is detected early - can't remember where I saw the numbers, but stuff like colon cancer diagnosis rates are way down this year, which will translate into more deaths over the next few years.
Of course, but this is likely to get lost in the overall noise, unless COVID carries on at the same levels for some considerable time.

Also remember that cancer deaths mostly occur in the same population groups who are vulnerable to COVID --- there's a lot of spin (much of it from people who want to go to the pub) about "poor young mothers dying of breast cancer because we shut down the NHS to save granny" (of course, if the hospitals overflow, we don't save granny and the young mother doesn't get her screening either), but cancer is an old person's disease --- we just hear a lot more about the outliers at the left-hand side of the age distribution.
Something something hammer something something nail

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Tue Nov 24, 2020 7:18 pm

AMS wrote:
Tue Nov 24, 2020 6:46 pm
Don't know if this is covered in any of the linked articles, but some non-covid excess deaths will take a while to actually happen if they result from missed diagnoses. In particular, cancer outcomes are generally better when it is detected early - can't remember where I saw the numbers, but stuff like colon cancer diagnosis rates are way down this year, which will translate into more deaths over the next few years.
Yes in the ONS analysis

Lots of competing effecte

https://www.ons.gov.uk/news/statementsa ... dmorbidity
Over the whole period presented from March 2020 to more than five years from now, there are approximately 25,000 excess deaths resulting from social distancing and economic impacts in the chosen scenario.

The negative health impacts of social distancing amount to a loss of 88,000 QALYs up to March 2021, so mortality improves over this period but health worsens – the main causes of this are increases in: depression; musculoskeletal disorders, resulting from more people working from home without suitable equipment; and domestic abuse.
Have you considered stupidity as an explanation

User avatar
sTeamTraen
Dorkwood
Posts: 1211
Joined: Mon Nov 11, 2019 4:24 pm
Location: Palma de Mallorca, Spain

Re: COVID-19

Post by sTeamTraen » Wed Nov 25, 2020 6:44 pm

The UK reported 18213 new cases and 696 deaths today. The 7-day moving averages are 18095 and 466. On the first day of the lockdown they were 22397 and 295. It's not clear to me why, if introducing the lockdown made sense on 5 November, lifting it on 2 December will do so.

From an objective point of view, this is based more on the cases than the deaths. If we imagine an infection-to-death time of 3 months, and cases fell to zero after a month, there would be two months worth of deaths baked in but nothing we could do about it; so the unlocking has to be based on cases, even though journalists and many ordinary people will focus on "Prime Minister, how can you ease the lockdown when more people are dying now than before it started". But of course, the current level of cases means that the deaths will continue at the current level as well.

However, given that cases have been fairly constant (around 20,000-25,000) for the past 4-5 weeks, and the actual time from infection to death seems to be a lot less than that, I wonder why deaths are going up so much now. Were/are there many undiagnosed cases, or is the second wave just now finding its way through to older people? Does anyone have recent age profile heat maps?
Something something hammer something something nail

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Wed Nov 25, 2020 7:37 pm

sTeamTraen wrote:
Wed Nov 25, 2020 6:44 pm
The UK reported 18213 new cases and 696 deaths today. The 7-day moving averages are 18095 and 466. On the first day of the lockdown they were 22397 and 295. It's not clear to me why, if introducing the lockdown made sense on 5 November, lifting it on 2 December will do so.

From an objective point of view, this is based more on the cases than the deaths. If we imagine an infection-to-death time of 3 months, and cases fell to zero after a month, there would be two months worth of deaths baked in but nothing we could do about it; so the unlocking has to be based on cases, even though journalists and many ordinary people will focus on "Prime Minister, how can you ease the lockdown when more people are dying now than before it started". But of course, the current level of cases means that the deaths will continue at the current level as well.

However, given that cases have been fairly constant (around 20,000-25,000) for the past 4-5 weeks, and the actual time from infection to death seems to be a lot less than that, I wonder why deaths are going up so much now. Were/are there many undiagnosed cases, or is the second wave just now finding its way through to older people? Does anyone have recent age profile heat maps?
We know that about 88% of deaths occur within 28 days of a positive test.
Have you considered stupidity as an explanation

KAJ
Stargoon
Posts: 93
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Wed Nov 25, 2020 9:00 pm

jimbob wrote:
Wed Nov 25, 2020 7:37 pm
sTeamTraen wrote:
Wed Nov 25, 2020 6:44 pm
The UK reported 18213 new cases and 696 deaths today. The 7-day moving averages are 18095 and 466. On the first day of the lockdown they were 22397 and 295. It's not clear to me why, if introducing the lockdown made sense on 5 November, lifting it on 2 December will do so.

From an objective point of view, this is based more on the cases than the deaths. If we imagine an infection-to-death time of 3 months, and cases fell to zero after a month, there would be two months worth of deaths baked in but nothing we could do about it; so the unlocking has to be based on cases, even though journalists and many ordinary people will focus on "Prime Minister, how can you ease the lockdown when more people are dying now than before it started". But of course, the current level of cases means that the deaths will continue at the current level as well.

However, given that cases have been fairly constant (around 20,000-25,000) for the past 4-5 weeks, and the actual time from infection to death seems to be a lot less than that, I wonder why deaths are going up so much now. Were/are there many undiagnosed cases, or is the second wave just now finding its way through to older people? Does anyone have recent age profile heat maps?
We know that about 88% of deaths occur within 28 days of a positive test.
This impinges on lag and delays which has become a hobby horse of mine, so beware a loser length post!

@jimbob: The death data at data.gov.uk is ...
Number of deaths of people who had had a positive test result for COVID-19 and died within 28 days of the first positive test.
... so 100% of those deaths occur within 28 days of a positive test.

With respect to "...cases have been fairly constant (around 20,000-25,000) for the past 4-5 weeks..."
this is the last couple of months data, cases v. specimen date. The lines are fitted values, the smoother line without the weekday() term to give a better idea of trend.
Cases.png
Cases.png (19.57 KiB) Viewed 105 times
As discussed earlier, delays between positive test and publishing on data.gov.uk can cause an apparent decline at recent dates. In the graph above I've omitted the last few data to mitigate that. In addition, to estimate those delays, since 17 Nov I've been saving the daily published data so I can see how long it takes for the numbers for a given specimen date to stabilise. Here's the data so far:
cCases.png
cCases.png (53.6 KiB) Viewed 105 times
Note the non-linear Y-scale to emphasise the later changes. It's clear that a very large proportion of cases are published within 5 days of the specimen date, suggesting that the decline in cases shown above is real.

Turning to "...4-5 weeks, and the actual time from infection to death seems to be a lot less than that..."
As pointed out above, published deaths are ""within 28 days of the first positive test" so the actual time between a case identification and death cannot be more than that. The CDC Interim Clinical Guidance for Management of Patients with Confirmed Coronavirus Disease (COVID-19) says...
Among patients in multiple early studies from Wuhan, China who had severe COVID-19 illness, the median time from their onset of illness to the time they experienced dyspnea was 5–8 days; the median time from onset of illness to acute respiratory distress syndrome (ARDS) was 8–12 days; and the median time from onset of illness to ICU admission was 9.5–12 days.
... so the time can be as little a few days.

But that's the time between actual positive specimen date and actual date of death. We're using published data. Similar to the case data above I've been saving published data.gov.uk death data. Here's the data so far:
cDeaths.png
cDeaths.png (65.45 KiB) Viewed 105 times
It looks as if there's at least a couple of weeks after a date before essentially all deaths on that date are reported. That's an additional lag which should be considered when considering how changes in deaths follow changes in cases.

User avatar
shpalman
After Pie
Posts: 2290
Joined: Mon Nov 11, 2019 12:53 pm
Location: One step beyond

Re: COVID-19

Post by shpalman » Wed Nov 25, 2020 9:15 pm

The UK graph of cases by specimen date now tends to have the last five days in grey to highlight that they are incomplete.

Regarding the reported case numbers, yesterday they were weirdly low so today might have seen some cases which took one day more to come through. This was definitely the case with Northern Ireland, whose numbers are however too small to make such a big difference to the whole of the UK.
sports-covid.png
sports-covid.png (139.45 KiB) Viewed 99 times
molto tricky

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Wed Nov 25, 2020 9:24 pm

sTeamTraen wrote:
Wed Nov 25, 2020 6:44 pm
The UK reported 18213 new cases and 696 deaths today. The 7-day moving averages are 18095 and 466. On the first day of the lockdown they were 22397 and 295. It's not clear to me why, if introducing the lockdown made sense on 5 November, lifting it on 2 December will do so.

From an objective point of view, this is based more on the cases than the deaths. If we imagine an infection-to-death time of 3 months, and cases fell to zero after a month, there would be two months worth of deaths baked in but nothing we could do about it; so the unlocking has to be based on cases, even though journalists and many ordinary people will focus on "Prime Minister, how can you ease the lockdown when more people are dying now than before it started". But of course, the current level of cases means that the deaths will continue at the current level as well.

However, given that cases have been fairly constant (around 20,000-25,000) for the past 4-5 weeks, and the actual time from infection to death seems to be a lot less than that, I wonder why deaths are going up so much now. Were/are there many undiagnosed cases, or is the second wave just now finding its way through to older people? Does anyone have recent age profile heat maps?
I haven't got a nice heatmap, but have just made this for the age profiles from the WK46 COVID Weekly registrations tab of the ONS data
AgeProfilesWk46.PNG
AgeProfilesWk46.PNG (39.17 KiB) Viewed 98 times
The heatmap was pretty rubbish to read - I haven't thought what colourscale to use - probably a logarithmic one - or maybe some weird percentile one
AgeProfilesWk46Heatmap.PNG
AgeProfilesWk46Heatmap.PNG (55.86 KiB) Viewed 95 times
Have you considered stupidity as an explanation

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Wed Nov 25, 2020 9:34 pm

KAJ wrote:
Wed Nov 25, 2020 9:00 pm
jimbob wrote:
Wed Nov 25, 2020 7:37 pm


We know that about 88% of deaths occur within 28 days of a positive test.
This impinges on lag and delays which has become a hobby horse of mine, so beware a loser length post!

@jimbob: The death data at data.gov.uk is ...
Number of deaths of people who had had a positive test result for COVID-19 and died within 28 days of the first positive test.
... so 100% of those deaths occur within 28 days of a positive test.
Indeed - that was what I was alluding to in a rather oblique way
Have you considered stupidity as an explanation

User avatar
jimbob
After Pie
Posts: 1838
Joined: Mon Nov 11, 2019 4:04 pm
Location: High Peak/Manchester

Re: COVID-19

Post by jimbob » Wed Nov 25, 2020 9:51 pm

A slightly less unclear heatmap in log5 bands
AgeProfilesWk46Heatmap2.PNG
AgeProfilesWk46Heatmap2.PNG (41.06 KiB) Viewed 81 times
Have you considered stupidity as an explanation

KAJ
Stargoon
Posts: 93
Joined: Thu Nov 14, 2019 5:05 pm
Location: UK

Re: COVID-19

Post by KAJ » Wed Nov 25, 2020 10:23 pm

shpalman wrote:
Wed Nov 25, 2020 9:15 pm
The UK graph of cases by specimen date now tends to have the last five days in grey to highlight that they are incomplete.
Thanks, I hadn't noticed that. I see they've done the same with deaths by date of death, where i think the incomplete period is longer. When I get time I'll adjust my R code to include but flag and zero-weight recent data.

I like the XKCD. I know I'm torturing the data, but I can't resist :ugeek:

Post Reply